Discovery and Investigation of Ammonia Decomposition Catalysts

Liquid ammonia can be used as a hydrogen transportation and generation source for use in PEM fuel cells. Current Ru catalysts for ammonia decomposition contain high loadings of Ru and require reaction temperatures at or above 550°C to attain equilibrium conversion. For on-site hydrogen generation, it is of interest to combine hydrogen generation from ammonia decomposition directly with PEM fuel cells. For this occur, operating temperatures need to be considerably lowered and effluent concentrations of ammonia need to be minimized to avoid poisoning of the membrane. Therefore, it is of interest to develop a low-cost catalyst that exhibits high activity at temperatures at or below 450°C. Prior work from our group discovered the use of supported hollandite (KRu4O8) exhibited excellent low temperature ammonia decomposition activity. This work further investigates under what conditions and synthesis parameters the hollandite structure can form, and further delves into the working state of the catalyst before exposure to ammonia. Here we show that the hollandite is a sacrificial structure that forms metallic Ru in various particle sizes depending on the H2 reduction temperature. Additionally, we compare these mixed metal Ru oxide catalysts with K promoted Ru catalysts synthesized via strong electrostatic adsorption (SEA). Next, we report the synthesis and high throughput catalytic screening of K promoted Ru based catalysts that have one of 31 additional metals incorporated, at three different Ru and secondary metal weight loadings. The Ru weight loading varied from 3 wt% to 1 wt%. In total, over 100 catalysts (including duplicates) were synthesized via incipient wetness impregnation method and screened for ammonia decomposition activity using a 16-channel parallel plug flow reactor. Fourier transform infrared (FTIR) imaging was used to analyze all 16 effluent streams in parallel in under two minutes. At 300°C, catalysts containing Mg, Ca, Sr, Sc, Y, Zr, Hf, Ta, Rh and Ir with 3wt% Ru were found to have excellent ammonia decomposition activity compared a K promoted 4wt% Ru catalyst that was previously optimized by our group. Catalysts containing 1 wt% Ru and 3 wt% Sc, Y, Zr or Hf were found to outperform the K promoted 4wt% Ru catalyst at the same reaction conditions. Many of these catalyst combinations reported here have not been reported for ammonia decomposition previously. Further insight into Sr and Fe containing catalysts were further investigated for their turnover frequency (TOF), apparent activation energy, H2 uptake, and through CO adsorption to understand mechanistically the difference between the two different kinds of catalysts. Additional insight into the working of the catalysts were investigated through XRD phase identification and profile fitting to determine how the different Ru species present, crystallite size and secondary metal influenced the ammonia decomposition activity. A machine learning algorithm was developed to extract the activity descriptors and elemental characteristics that are responsible for ammonia decomposition activity at different operating temperatures. We demonstrate the application of a random forest machine learning algorithm to high-throughput experimental data to increase understanding of catalyst behaviour through knowledge extraction and to guide catalyst discovery through predictions. The knowledge extracted from this material agnostic machine learning algorithm can be used to design a second iteration of catalysts, where features that contributed to the greatest change in activity were accentuated. Additionally, this information can be further applied to the design of ammonia synthesis catalysts at ambient pressures